Introduce parameter to fix deepspeed crash for RNNS

[SeanNaren]

What does this PR do?

Fixes #9394

By introducing partition_module, the user is able to turn of the explicit conversion of the entire module. This leaves DeepSpeed to do the conversion itself.

The reason in most cases its best to do it upfront, is because we can control the parameters passed to the initialisation + if the user defines layers in configure_sharded_model we can also ensure all other layers are converted (Which is a requirement).

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[SeanNaren]

I did look into the possibility of automating this. However within DeepSpeed the conversion happens without passing necessary information to do the conversion: https://github.com/microsoft/DeepSpeed/blob/master/deepspeed/runtime/zero/stage3.py#L663

Thus we need to do this beforehand to ensure we setup correctly.

[tchaton]

LGMT !